This year, the Laboratory of Computational Biology procured significant GPU resources to the LoBoS cluster, in the form of 24 nodes equipped with dual Nvidia K40 GPUs (48 GPUs total) and interconnected via InfiniBand. This cluster will substantially increase the amount of computing power available to the lab, as CHARMM's DOMDEC GPU code has shown itself able to take advantage of multiple GPUs on multiple nodes. This will allow the lab to explore microsecond-level time scales of fairly large simulations. Furthermore, the lab has been planning for a new computer room, to be located on the first floor of the NIH building 12 data center, which will replace the existing facility at Fishers Lane. Since the lab is being moved back to the NIH main campus and NIH will not maintain space at 5635 Fishers Lane, this new machine room is necessary to continue the high performance computing efforts. Design work is being done in collaboration with the Space and Facilities Section of the Intramural Administrative Management Branch, the Center for Information Technology, and the Office of Research Facilities. Molecular simulation and modeling software packages are the vehicle for computational research and experiment. Implementation of new methods and options is the key to facilitate cutting edge researches. In recent years, this lab has developed a series new compuatational methods, such as the self-guided Langevin dynamics for efficient conformational searching and sampling, the isotropic periodic sum method for accurate and efficient calculation of long-range interactions, and the map-based modeling tool, EMAP, for electron microscropy studies. Implementation of these new methods enables researchers to tackle difficult problems. These methods have been implemented into CHARMM to expand its capability in molecular simulation, conformational search, and structure prediction. These methods are all available in CHARMM version 40. These methods are also been implemented into another widely used simulation package, AMBER, to extend the user scope to access these methods. The SGLD, IPS, and EMAP methods are available in AMBER version 14. Next generation software methods Our fast polarization and multipole methods have been implemented in the CHARMM, TINKER and OpenMM software packages to make calculations using the AMOEBA force field much faster. We have also used our spherical harmonic multipole algorithm to implement the SSMP water model in CHARMM; unlike its SSDQO predecessor, SSMP can be run in concert with existing force fields, which greatly expands its audience. We have also developed and implemented dispersion PME methods to allow non-covalent interactions to be treated effectively with infinite cutoffs; this development is currently being used to reparameterize existing force fields and to build new polarizable methodology. MPID polarizable force field The CHARMM MPOLE module was further developed for more general and flexible support of polarizable force fields besides the AMOEBA force field. Implementations include anisotropic atomic polarizability, alternative Thole screening functions, dispersion particle mesh Ewald (PME), and the mixing of polarizable and non-polarizable sites. Such software development allows us to transfer the Drude polarizable force field from its original Drude oscillator model into the permanent multipole and induced dipole model. We have shown that a direct mapping reproduces condensed phase properties of liquids, crystals and interfaces without any need to reparametrize. This indicates that the two major approaches to include polarizability - the classical Drude oscillator model and the point induced dipole model - are different representations of the same physical model. The resulting MPID polarizable force field is computationally efficient and is ready to be used in MD simulations and QM/MM calculations. Our simulations with the MPID force field also serve as a validation of the OPT3 analytical algorithm for induced dipoles developed in the lab. A genetic algorithm for optimizing force field parameters based on expensive QM calculations In molecular dynamics simulation, force field parameters have a large influence on the accuracy of the final calculated predictions. To automate this difficult and tedious process, a genetic algorithm has been applied determine parameters to reproduce methanol cluster data generated using high level quantum mechanics calculations. This approach yields parameters which are able to accurately reproduce experimental bulk properties such a density and heat of vaporization, despite the fact that only QM calculations were used to fit the parameters.